This is a protocol wrapper that enforces three specific behaviors: how to receive code review feedback without performative agreement, when to dispatch a code-reviewer subagent (after each task in subagent workflows, before merges), and most importantly, verification gates that block any completion claims until you've actually run the commands and seen the output. The iron law here is no "tests should pass" or "seems to work" without fresh evidence. It's opinionated about stopping false confidence in AI coding workflows, which honestly feels necessary given how often LLMs claim success without verification. Best suited for subagent-driven development where you need forcing functions against premature task completion.
npx skills add https://github.com/secondsky/claude-skills --skill code-review